ABSTRACT
This chapter explores a weights-and-structure-determination (WASD) algorithm is presented for a power-activation neuronet to solve monthly time series learning and prediction problems. It discusses a power-activation feed-forward neuronetfeed-forward neuronet with the three-layer structure is constructed for time seriestime series learning and prediction. The chapter investigates time seriestime series learning and prediction via neurone. It also explores the abilities of neuronets in monthly time seriestime series learning and prediction, a power-activation neurone. Compared with statistics-based prediction techniques, the following several unique characteristics of neuronets make them valuable and attractive for time seriestime series prediction The chapter analyses a simple and effective data preprocessingdata preprocessing approach to adjust the time seriestime series which may contain various patterns, such as trend and seasonality. Numerical studies further substantiate the superiority of the power-activation neuronet equipped with the WASD algorithm to predict monthly time series.
